Network patterns of legislative collaboration in twenty parliaments

In recent years, the ties that Members of Parliament (MPs) create by cosponsoring legislation together have attracted interest from scholars adopting a network approach to lawmaking. is brief note expands the empirical base of these studies by introducing a dataset of 150 bill cosponsorship networks that cover 27 parliamentary chambers from 19 European countries, plus Israel. The data show the extent of partisanship expressed by MPs through their propensity to cosponsor bills within and across party lines, in several different parliamentary systems.

1 Our search also returned visual explorations of bill cosponsorship in the lower houses of the French (Coulmont, 2011) and Czech (Gregor, 2013) parliaments, and similar research on Korean legislators (Lee & Yoon, 2014). 2 We focused our attention on private bills, defined as laws initiated by one or more MPs that become binding if they make it through the legislative process of their country of introduction. This definition is compatible with theoretical assumptions on how MPs signal their positions to their constituents or to third parties, and is comparable across countries. 3 In the unique case of Israel, we also used an open data portal maintained by an unaffiliated third party, Open Knesset, by The Public Knowledge Workshop: https://oknesset.org/.  Table 1 summarises the data that we managed to collect, which cover 20 countries and 27 parliamentary chambers, over a total of 558 years and 150 legislatures, understood as periods between two nationwide legislative elections. The sample contains a mix of unicameral and bicameral parliamentary systems, including three federal regimes (Austria, Belgium, and Switzerland). The R code (R Core Team, 2015) used to collect the data and assemble the cosponsorship networks is available at https://github.com/briatte/parlnet, along with detailed replication instructions.
Using the same parliamentary sources as we used for bills, we also retrieved as much information as possible on the individual legislators who nominally sponsored the bills. The variables collected across all countries include sponsor age, sex, and parliamentary career information (time in office, constituency, committee membership, and party affiliation), for a total of approximately 18,000 MPs who appeared on at least one cosponsored bill. To further characterize the positions of bill sponsors relative to each other, we also proceeded to match their party affiliations with an indication of where the party sits on a standardized Left/Right scale. In order to do so, we used the scores available in the latest edition of the ParlGov database (Döring & Manow, 2014), which are time-invariant scores computed as  Figure 1 shows one of the cosponsorship networks that can be constructed from the data we collected, using force-directed placement (Fruchterman & Reingold, 1991). The network, which shows bill cosponsorship ties in the ongoing legislature of the unicameral parliament of Sweden, is a onemode projection of the b × a two-mode matrix, where b denotes bills and a denotes their sponsors, that connects the first author of each bill to all other sponsors on that bill. The resulting adjacency matrix A ij of directed ties between MPs (i, j) is asymmetric and contains no self-loops.
In order to further explore the collaborations that take place in legislative cosponsorship networks, we also built interactive versions of the same graphs, which allow the user to explore the ego networks of specific MPs. These visualizations, an example of which is shown in Figure 2, are available online at http://f.briatte.org/parlviz/. Last, because legislative cosponsorship networks are based on ties that represent one or more than shared bill(s) between two MPs, we computed several measures to weight their edges accordingly. These measures (raw cosponsorship counts, the weighted quantity of bills cosponsored, and the weighted propensity to cosponsor) are taken from existing studies of legislative cosponsorship in the U.S. Congress (Fowler, 2006a;Gross et al., 2012), and are documented in full in the appendix to this note.
As illustrated in Figure 1 and as visible in the interactive visualizations previously mentioned, all of the 150 observed networks clearly show the influence of party affiliations over decisions to cosponsor bills. Using these data, the extent of partisanship expressed by MPs through their propensity to cosponsor bills within and across party lines might be measured through different methods: several studies of the U.S. Congress (Zhang et al., 2008;Waugh et al., 2009;Moody & Mucha, 2013) use the modularity network statistic (Newman & Girvan, 2004;Leicht & Newman, 2008) to that effect, but the data are also amenable to other estimation methods, such as exponential random graph models (Cranmer & Desmarais, 2011;Snijders, 2011). Such measures should confirm that, as Sartori (1976Sartori ( /2005 observed several decades ago, the distribution of power between political parties can take many different forms in highly competitive electoral environments, as "the fragmentation of the party system can reflect either a situation of segmentation or a situation of polarization, i.e. of ideological distance" (p. 111).
The levels of party polarization shown in the networks under study represent only one of many possible ways to explore the individual and institutional determinants that govern over the decisions of MPs to cosponsor each other's bills. In similar fashion to Moody and Mucha (2013), we therefore hope that the data presented in this note, might serve as an introduction to a complex empirical puzzle, extended to a set of country cases that allow for comparative inquiry, and supported by interactive network visualizations.